Vercel Adds Wan 3.0 to AI Gateway for Longer Video Generation.

Vercel is bringing longer, higher-resolution AI video generation to its developer platform. The company has added Alibaba’s Wan 3.0 model to AI Gateway, giving applications one API route for generating video from text, images, existing video and audio references.

The model is available now through AI Gateway under the ID alibaba/wan-v3.0-video. It can produce clips lasting up to 30 seconds at 30 frames per second, with output resolutions of 480p, 720p or 1080p. Generated videos can also include synchronized audio.

That combination makes Wan 3.0 a notable upgrade for builders creating automated marketing assets, product demonstrations, social video tools or user-generated media features. Instead of connecting separate models for different input types, developers can send several kinds of generation requests to one model.

Wan 3.0 combines four previously fragmented capabilities:

• Text-to-video, for generating footage from a written prompt
• Image-to-video, for animating a still image
• First- and last-frame conditioning, for controlling how a clip starts and ends
• Reference-based generation, using images, video or audio as guidance

The practical change is not only model quality or output length. Consolidating these workflows under one model ID can reduce routing logic inside an application. A product no longer needs to select separate Wan variants depending on whether a user begins with text or supplies reference media.

Vercel says the previous Wan 2.7 integration required distinct -t2v and -r2v model IDs for text-to-video and reference-to-video tasks. It was also limited to 15-second clips at 24 fps. Wan 3.0 doubles the maximum duration and increases the frame rate to 30 fps, while adding support for output as high as 1080p.

For small teams, those changes could make AI video features easier to prototype without building a dedicated media-generation backend. AI Gateway provides a common access layer across AI providers, alongside infrastructure for areas such as rate limiting, caching and observability. Developers can therefore add Wan 3.0 while keeping model access and operational monitoring within the same gateway used by the rest of an AI application.

Video generation can take substantially longer than a typical text-model response. Keeping an HTTP connection open throughout that process is fragile, particularly in serverless applications where requests may be constrained by execution limits or interrupted by a user closing the page.

Wan 3.0 supports Vercel AI Gateway’s asynchronous video generation. An application can begin rendering, release the original request and receive a webhook event when the job finishes.

A basic request using Vercel’s AI SDK follows this pattern:

import { experimental_generateVideo as generateVideo } from 'ai';

const { videos } = await generateVideo({
  model: 'alibaba/wan-v3.0-video',
  prompt: 'A paper lantern drifting above a city at night',
  webhook: 'https://example.com/api/video-complete'
});

This background-job approach matters for production workflows. A storefront could generate a product clip after an image upload, for example, then notify the customer when it is ready. A content tool could submit several jobs without forcing its interface to wait on each render.

Vercel also lists start-and-status requests and short polling as ways to monitor asynchronous video jobs. Webhooks will generally be the cleaner option for automated workflows, while polling may be useful when a client needs to show live progress or cannot expose a webhook endpoint.

There are still implementation details for teams to evaluate, including generation cost, storage, content review and the handling of uploaded reference media. Vercel’s announcement does not specify Wan 3.0 pricing or plan requirements, so builders should verify current terms before designing a high-volume feature around it.

What is Vercel Wan 3.0?

Wan 3.0 is Alibaba’s AI video model, made available to developers through Vercel AI Gateway. It supports text-to-video, image-to-video, frame conditioning and reference-based generation, with synchronized audio and clips of up to 30 seconds.

How can developers access Wan 3.0 on Vercel?

Developers can call the model through AI Gateway using the model ID alibaba/wan-v3.0-video. It supports asynchronous rendering, allowing an application to receive a webhook or check the job’s status instead of holding an HTTP request open.

How does Wan 3.0 compare with Wan 2.7?

Wan 2.7 used separate model IDs for text-to-video and reference-to-video generation, and produced clips of up to 15 seconds at 24 fps. Wan 3.0 unifies those workflows and supports clips up to 30 seconds at 30 fps, with resolutions reaching 1080p.